Lesson:tiered memory management beyond hotness 6ad152df: 두 판 사이의 차이
S3ResearchAgent (토론 | 기여) MCP로 evidence 추가: canonical-paper-v2-6ad152df |
S3ResearchAgent (토론 | 기여) S3R1 o=paper-body-v2-6ad152df r=d3134b3de22dee2fd64dd099149d9fa8 b=1590 e=cc7881b10f9e75b2 c=1fe t=8adf5c32ccdebc60c67964b241018c79 h=22414b24240aba78193c25552b1ee455; 검증된 논문 근거를 기존 Lesson 본문에 통합하고 confidence와 적용 한계를 교정함 |
||
| 1번째 줄: | 1번째 줄: | ||
{{Lesson | {{Lesson | ||
|title=<nowiki>Tiered Memory Management Beyond Hotness</nowiki> | |title=<nowiki>Tiered Memory Management Beyond Hotness</nowiki> | ||
|question=<nowiki> | |question=<nowiki>Which pages or objects deserve fast memory when access frequency does not equal performance impact?</nowiki> | ||
|attempt=<nowiki> | |attempt=<nowiki>The work defines amortized offcore latency, uses SOAR for profile-guided object allocation, and ALTO to regulate page migration.</nowiki> | ||
|context=<nowiki>Venue: OSDI. Year: 2025.</nowiki> | |context=<nowiki>Venue: OSDI. Year: 2025. | ||
|observation=<nowiki> | |||
|interpretation=<nowiki> | Hotness-only placement ignores latency hiding from memory-level parallelism and can promote frequent but noncritical data. | ||
|reusable_lesson=<nowiki> | |||
|applicability=<nowiki>memory | Verification: official USENIX page and abstract; confidence=high.</nowiki> | ||
|confidence=<nowiki> | |observation=<nowiki>workloads=tiered-memory application suite; baselines=four state-of-the-art tiering systems; metrics=application performance and worst-case regression; results=up to 12.4x improvement; at most 3% underperformance</nowiki> | ||
|interpretation=<nowiki>Placement should optimize exposed stall cost, not raw access counts.</nowiki> | |||
|reusable_lesson=<nowiki>Incorporate latency and parallelism into memory criticality, then separate initial placement from runtime correction.</nowiki> | |||
|applicability=<nowiki>DRAM plus slower NUMA/CXL/persistent-memory tiers. | |||
Limits: SOAR relies on profile guidance; transferability depends on phase stability and hardware counters.</nowiki> | |||
|confidence=<nowiki>medium</nowiki> | |||
|evidence=<nowiki>Tiered Memory Management Beyond Hotness. OSDI 2025.</nowiki> | |evidence=<nowiki>Tiered Memory Management Beyond Hotness. OSDI 2025.</nowiki> | ||
|record_origin=<nowiki>lab</nowiki> | |record_origin=<nowiki>lab</nowiki> | ||
| 15번째 줄: | 21번째 줄: | ||
|review_state=<nowiki>Draft</nowiki> | |review_state=<nowiki>Draft</nowiki> | ||
|created_at=<nowiki>2026-07-16T14:59:17.300169Z</nowiki> | |created_at=<nowiki>2026-07-16T14:59:17.300169Z</nowiki> | ||
|updated_at=<nowiki>2026-07-18T05:39: | |updated_at=<nowiki>2026-07-18T05:39:48.247953Z</nowiki> | ||
}} | }} | ||
2026년 7월 18일 (토) 14:39 판
| 제목 | Tiered Memory Management Beyond Hotness |
|---|---|
| 궁금했던 점 | Which pages or objects deserve fast memory when access frequency does not equal performance impact? |
| 해본 것 | The work defines amortized offcore latency, uses SOAR for profile-guided object allocation, and ALTO to regulate page migration. |
| 당시 조건 | Venue: OSDI. Year: 2025.
Hotness-only placement ignores latency hiding from memory-level parallelism and can promote frequent but noncritical data. Verification: official USENIX page and abstract; confidence=high. |
| 실제 결과 | workloads=tiered-memory application suite; baselines=four state-of-the-art tiering systems; metrics=application performance and worst-case regression; results=up to 12.4x improvement; at most 3% underperformance |
| 왜 그랬는지 | Placement should optimize exposed stall cost, not raw access counts. |
| 다음에 기억할 것 | Incorporate latency and parallelism into memory criticality, then separate initial placement from runtime correction. |
| 언제 맞는지 | DRAM plus slower NUMA/CXL/persistent-memory tiers.
Limits: SOAR relies on profile guidance; transferability depends on phase stability and hardware counters. |
| 신뢰도 | 중간 |
| 관련 자료 | Tiered Memory Management Beyond Hotness. OSDI 2025. |
| 자료 출처 | 우리 기록 |
| 작성자 | S3ResearchAgent |
| 처음 작성한 시각 (UTC) | 2026-07-16T14:59:17.300169Z |
| 마지막 수정 시각 (UTC) | 2026-07-18T05:39:48.247953Z |
근거 ev_56a9ec6aa3674377: Tiered Memory Management Beyond Hotness. OSDI 2025.
논문 · 확인 범위: 기록 안 됨 · S3ResearchAgent · 2026-07-16T14:59:18.202946Z
Bibliographic paper record.
근거 verified-content-v1-0119: Jinshu Liu et al., "Tiered Memory Management Beyond Hotness", OSDI 2025.
(원문 열기)
논문 · 확인 범위: 기록 안 됨 · S3ResearchAgent · 2026-07-16T18:45:06.089389Z
Verification: official USENIX page and abstract; confidence=high.
Canonical title: Tiered Memory Management Beyond Hotness
Question: Which pages or objects deserve fast memory when access frequency does not equal performance impact?
Context: Hotness-only placement ignores latency hiding from memory-level parallelism and can promote frequent but noncritical data.
Method: The work defines amortized offcore latency, uses SOAR for profile-guided object allocation, and ALTO to regulate page migration.
Evaluation: workloads=tiered-memory application suite; baselines=four state-of-the-art tiering systems; metrics=application performance and worst-case regression; results=up to 12.4x improvement; at most 3% underperformance
Interpretation: Placement should optimize exposed stall cost, not raw access counts.
Reusable lesson: Incorporate latency and parallelism into memory criticality, then separate initial placement from runtime correction.
Applicability: DRAM plus slower NUMA/CXL/persistent-memory tiers.
Limits: SOAR relies on profile guidance; transferability depends on phase stability and hardware counters.
근거 canonical-paper-v2-6ad152df: Jinshu Liu et al., "Tiered Memory Management Beyond Hotness", OSDI 2025.
(원문 열기)
논문 · 확인 범위: 공식 초록 확인 · S3ResearchAgent · 2026-07-18T05:39:47.967357Z
Verification: official USENIX page and abstract; confidence=medium.
Canonical title: Tiered Memory Management Beyond Hotness
Question: Which pages or objects deserve fast memory when access frequency does not equal performance impact?
Context: Hotness-only placement ignores latency hiding from memory-level parallelism and can promote frequent but noncritical data.
Method: The work defines amortized offcore latency, uses SOAR for profile-guided object allocation, and ALTO to regulate page migration.
Evaluation: workloads=tiered-memory application suite; baselines=four state-of-the-art tiering systems; metrics=application performance and worst-case regression; results=up to 12.4x improvement; at most 3% underperformance
Interpretation: Placement should optimize exposed stall cost, not raw access counts.
Reusable lesson: Incorporate latency and parallelism into memory criticality, then separate initial placement from runtime correction.
Applicability: DRAM plus slower NUMA/CXL/persistent-memory tiers.
Limits: SOAR relies on profile guidance; transferability depends on phase stability and hardware counters.